RNA-seq mini-project — Salmon → DESeq2 (Mouse, mm10)

This repository contains a reproducible RNA-seq differential expression pipeline based on Salmon, tximport, and DESeq2, applied to a mouse dataset (mm10, GENCODE).

The goal of this mini-project was to practice RNA-seq data analysis concepts in parallel with the Bioinformatics Methods for Transcriptomics course (Johns Hopkins University, Coursera), with a focus on:

• transcript-level quantification with Salmon
• gene-level summarization with tximport
• differential expression analysis with DESeq2
• writing a fully reproducible R script (no interactive-only analysis)

---

Data source and timeline

• Dataset: GSE83931 (mouse RNA-seq)
• Source: Dataset used in the course Bioinformatics Methods for Transcriptomics
  Johns Hopkins University — Coursera
• Organism: Mouse (mm10)
• Project duration: December 6–14, 2025

This project was conducted alongside the course, with the objective of implementing a complete RNA-seq workflow under realistic local computing constraints.

Due to hardware and memory limitations on a personal laptop, some tools presented in the course (e.g. STAR and other memory-intensive aligners) could not be run locally. To address this, the analysis was performed under Linux using Windows Subsystem for Linux (WSL), and a lightweight, transcript-level quantification approach (Salmon) was chosen as a technically appropriate and resource-efficient alternative.

---

Project structure

GSE83931_miniFASTQ/
├── data/
│   └── metadata/
│       └── samples.csv
├── reference/
│   └── gencode_mm10/
│       ├── gencode.vM23.transcripts.fa.gz
│       └── tx2gene_from_fasta.tsv
├── results/
│   ├── salmon/
│   └── DESeq2/
│       ├── DESeq2_results.csv
│       ├── DESeq2_results_clean.csv
│       ├── DESeq2_significant_padj0.1.csv
│       ├── dds.rds
│       ├── txi.rds
│       ├── sessionInfo.txt
│       └── run_log.txt
├── figures/
│   ├── PCA_plot.pdf
│   ├── MA_plot.pdf
│   └── Volcano_plot.pdf
└── scripts/
    └── RNAseq_Salmon_tximport_DESeq2.R

---

Experimental design

• Organism: Mouse (mm10)
• Quantification: Salmon
• Annotation: GENCODE vM23
• Conditions:
  • CTRL (control)
  • TREAT (treated)
• Contrast tested: TREAT vs CTRL

To reduce computational load and memory usage, a limited subset of samples was selected for analysis (1 male and 1 female at each time point). This choice enables local RNA-seq processing on a personal computer but reduces statistical power.

---

How to run the analysis

Run the pipeline from the project root:

cd ~/projects/GSE83931_miniFASTQ
Rscript scripts/RNAseq_Salmon_tximport_DESeq2.R

For maximum stability (recommended on laptops / WSL):

nohup Rscript scripts/RNAseq_Salmon_tximport_DESeq2.R > results/DESeq2/nohup.out 2>&1 &

---

Main outputs

• Differential expression (clean)
  results/DESeq2/DESeq2_results_clean.csv

• Significant genes (padj < 0.1)
  results/DESeq2/DESeq2_significant_padj0.1.csv

• Saved R objects

  • dds.rds
  • txi.rds

• Figures

  • PCA plot
  • MA plot
  • Volcano plot

• Reproducibility

  • sessionInfo.txt (R and package versions)
  • run_log.txt (full execution log)

---

Notes on reproducibility

• The pipeline rebuilds tx2gene automatically from the GENCODE FASTA if missing.
• Transcript identifiers are kept consistent with the Salmon index (GENCODE-style IDs).
• All steps are fully scripted; no results depend on interactive R history.

The analysis was designed to be reproducible on standard personal hardware with limited RAM.

Purpose of this project

This mini-project is part of a broader roadmap to:

• strengthen practical RNA-seq analysis skills
• build a portfolio of reproducible bioinformatics workflows
• prepare for advanced training and research in genomics and epigenetics